Modes of Apprehension, and Indicators thereof, in Visual Discrimination of Relative Mass

Full text

(1)

(2)

(3)

(4)

(5)

(6)

(7) .

(8)

(9)

(10) !"" #!$#.

(11)

(12) . #%&'()( #'*+'%,*&(% -.- -- %(+%,.

(13) .

(14)

(15)

(16) !"##$ %&##'

(17) (

(18) '

(19)

(20)

(21) ' )

(22)

(23) *+

(24) ,

(25) - (* .

(26) *"##$*/

(27)

(28) '.

(29)

(30)

(31) '

(32)

(33) '0 / *. . *

(34)

(35)

(36)

(37)

(38)

(39) 1"* *23 $45$ 1164!#" * )

(40) ' '

(41)

(42) ,

(43) (

(44)

(45) ,

(46) 7

(47)

(48)

(49)

(50) *8

(51)

(52) ( '

(53) ,

(54) ,'

(55)

(56) (

(57) '

(58) '

(59)

(60) *+

(61)

(62) , ' 9:

(63) ( :;

(64) 9 ;

(65)

(66) '

(67)

(68)

(69)

(70) ' < =

(71)

(72) '

(73) ' .

(74)

(75) *+

(76)

(77)

(78)

(79)

(80) , '

(81) 0

(82) >

(83) 9"###;,

(84)

(85) (

(86) ' ,

(87) <9?

(88) $!! $4$;* +

(89)

(90) '

(91) , ( (

(92)

(93)

(94) '

(95) ' ,

(96)

(97) (

(98) 7 *+ '

(99)

(100)

(101)

(102) . ,

(103) <, '

(104) (

(105) *8

(106) ,

(107)

(108) '

(109)

(110)

(111) ' (

(112)

(113) '

(114) 90

(115) *"###;*

(116) .

(117)

(118) ,

(119) & @

(120) '

(121) '

(122)

(123)

(124)

(125) '

(126)

(127) * ?

(128)

(129)

(130)

(131)

(132)

(133) ' < =

(134) *

(135) ((

(136)

(137) ' (

(138) ' ( (

(139) '

(140)

(141) * 8

(142) ,

(143)

(144)

(145)

(146) *

(147)

(148) '

(149)

(150)

(151)

(152) '

(153) ''

(154)

(155) '

(156)

(157)

(158) * (

(159)

(160)

(161)

(162) , ,

(163)

(164)

(165) '

(166)

(167) ' (

(168) (

(169)

(170)

(171)

(172) ( '

(173) .

(174)

(175) ,

(176)

(177) ( ' ,

(178) <*

(179) /

(180)

(181) '

(182)

(183)

(184)

(185) (

(186)

(187) ( !. "

(188) #

(189) $

(190)

(191) #%

(192) &'(()# #*+)',( # A .

(193) "##$ 3 !1"$#%# 23$45$ 1164!#" & &&& #5" 69. &BB *<*B

(194) C D & &&& #5" 6;.

(195) List of papers. This thesis is based on the following studies, which will be referred to in the text by their Roman numerals: I. Runeson, S., & Andersson, I. E. K. (2007). Achievement of specificational information usage with true and false feedback in learning a visual relative-mass discrimination task. Journal of Experimental Psychology: Human Perception and Performance, 33, 163-182.. II. Andersson, I. E. K., & Runeson, S. (2008). Realism of confidence, modes of apprehension, and variable-use in visual discrimination of relative mass. Ecological Psychology, 20, 1-30.. III. Andersson, I. E. K., Kreegipuu, K., Allik, J., & Runeson, S. (2009). Phenomenological reports, event-related potentials, and response times pertaining to mode of apprehension in visual discrimination of relative mass. (manuscript). IV. Andersson, I. E. K. (2009). Realism of confidence and phenomenological reports are not congruent indicators of modes of apprehension in visual discrimination of relative mass. Ecological Psychology, 21, 218-244..

(196)

(197) Contents. Introduction.....................................................................................................7 Direct perception ........................................................................................7 Perceptual errors and nonspecifying variables ......................................9 Perceptual learning ..............................................................................10 Indirect perception....................................................................................12 Modes of apprehension and the mode-transition model...........................14 What are the inferences in the inferential mode of apprehension? ......15 Mode transition: a shift from "thinking" to "perceiving" ....................16 Aims of the thesis..........................................................................................18 Method ..........................................................................................................19 Experimental paradigm ............................................................................19 Candidate informational variables .......................................................19 Equipment and simulations..................................................................22 Tentative indicators of mode of apprehension .........................................22 Variable usage .....................................................................................23 Realism of confidence .........................................................................24 Phenomenological mode reports..........................................................26 Event-related potentials and response times........................................28 Empirical studies...........................................................................................31 Study I ......................................................................................................31 Results .................................................................................................32 Study II.....................................................................................................35 Results .................................................................................................36 Study III ...................................................................................................39 Results .................................................................................................40 Study IV: Experiment 1............................................................................47 Results .................................................................................................48 Study IV: Experiment 2............................................................................49 Results .................................................................................................50 Study IV: Experiment 3............................................................................53 Results .................................................................................................53 Discussion .....................................................................................................55.

(198) Main findings and their limitations ..........................................................55 Variable usage and learning.................................................................55 Realism of confidence .........................................................................57 Mode reports........................................................................................62 Response times ....................................................................................63 Event-related potentials .......................................................................64 Mode reports and over/underconfidence scores combined..................65 Conclusions ..............................................................................................66 Direct perception and the inferential mode ..............................................68 Acknowledgements.......................................................................................73 References.....................................................................................................74.

(199) Introduction. “The constantly looming catastrophes of the intellect would be found more often developing into catastrophes of action were it not for the mellowing effect of the darker, more feeling-like and thus more dramatically convincing primordial (that is, perceptual) layers of cognitive adjustment.” (Brunswik, 1956, p. 93) "Merely probable information, clues or cues, is not as satisfying for the perceptual system as the achieving of clarity…but guessing does occur in highly complex situations and the individual may sometimes have to be content with it." (Gibson, 1966, p. 304). Perception is a fundamental function for all organisms. It allows organisms to be in contact with the environment and to adjust to environmental conditions. In addition, humans possess higher intellectual functions, which allow for elaborate handling of perceptually obtained information and of abstract entities. However, neither perception nor intellectual functioning always results in perfect adjustment to the environment, and both may improve with extended experience and practice. This thesis concerns perceptual skill acquisition, and more specifically an idea of perceptual skill acquisition as a transition from an inferential to a perceptual mode of apprehension (Runeson, Juslin, & Olsson, 2000). This mode-transition model of perceptual skill acquisition was formulated within the ecological direct-perception framework (Gibson, 1966, 1979; Michaels & Carello, 1981). Before describing the modes of apprehension and the mode-transition model, the ideas and concepts of the direct-perception approach are presented and contrasted to indirect approaches to perception.. Direct perception The direct-perception approach portrays perception as the extraction of information inherent in energy patterns that surrounds the organism (Gibson, 1966, 1979). The extraction of information occurs through smart mechanisms, which register apparently complex informational variables in the ambient stimulus flux, without prior detection and computation on simpler variables (Runeson, 1977; see also Michaels & Carello, 1981). Perception is said. 7.

(200) to be direct because unconscious inferences or other types of mental computations are superfluous due to the specificity of information. Information, in the Gibsonian sense, is a relation between energy (such as light, vibrations, etc.) and properties of the environment (such as surfaces, substances, etc.) (cf. Chemero, 2003). Energy fields are structured by sources in the environment. When described at the appropriate level, stimulation is thus not ambiguous with respect to environmental properties. Rather, it stands in a one-to-one relation to those properties, providing specificity of information. Informational specification is found in invariants, which are patterns of stimulation that co-vary with environmental properties, and that remain unchanged by circumstantial transformations over time or space (Gibson, 1966, 1979; Michaels & Carello, 1981; cf. Rogers, 2000). That is, spatio-temporal structures of energy provide specification of the environment. By detection and pick-up of specificational invariants, perceivers become informed about useful environmental properties, or about perceiver-relevant action possibilities (affordances) (Gibson, 1979; cf. Bingham, 2000). The specificity of information is granted by regularities that prevail in the world, that is, constraints (Runeson, 1988, 1989; Chemero, 2003). Constraints can be universal, such as general laws of optics and physics (e.g., laws of wave propagation), or local ecological constraints, such as the regular distribution of texture elements on surfaces. In addition, cultural conventions and rules can provide more confined constraints. Since specificity of information is depending on pertinent constraints that prevail, specificity will vary in degree of universality depending on the generality of constraints. The perceptual system may exploit any regularity (Warren, 2005), and it has evolved to take advantage of global and specific ecological constraints (Runeson, 1994). Individuals may also learn to take advantage of more local constraints, such as regularities in an experimental environment (e.g., Jacobs, Runeson, & Michaels, 2001). An instance of informational specificity, granted by constraints in the form of the laws of physics, is expressed by the principle of kinematic specification of dynamics (the KSD principle). The KSD principle states that kinematic patterns (motions) specify dynamic properties (causal factors such as force, mass, and friction) of moving objects (Runeson, 1977/1983; Runeson & Frykholm, 1983). As argued by Runeson (1994), it is often more action relevant to perceive dynamic properties such as mass of objects, than kinematic properties such as particular motions. Often, dynamic properties are lasting characteristics of objects, and by becoming informed about such properties, an observer may to some extent predict future movements of the objects involved, which facilitates future interaction with the objects.1 Thus, 1. However, this does imply that motions as such may not be relevant to observers, nor that motions as such are never perceived (cf. Study I and IV in this thesis).. 8.

(201) there is incentive to evolutionary development of perceptual systems that can respond to KSD-type of information. That observers can take advantage of KSD-type of information has been demonstrated in studies of visual perception of weight of lifted boxes, and of sex and identity of point-light walkers (Kozlowski & Cutting, 1977; Runeson & Frykholm, 1981, 1983). As will be discussed later, it has also been demonstrated in visual perception of relative mass of colliding objects (Cohen, 2006; Jacobs, Michaels, & Runeson, 2000; Jacobs, Runeson, & Michaels, 2001; Runeson et al., 2000; Runeson & Vedeler, 1993).. Perceptual errors and nonspecifying variables Generally, in the direct-perception view, perception is approached from the "perfectionist end" and perception is seen as "approximating perfection" (Runeson et al., 2000, p. 531), given that the observer is allowed to explore the optic array (for instance by moving around) (Gibson, 1966). Nevertheless, perceptual achievements are not always perfect. According to Gibson (1966), there are two reasons perception may fail; one is because adequate information is not being picked up due to deficient physiological processes, the other is because the available information is inadequate. Examples of misperceptions are mirages, and when a straight stick appears bent when part of it is submerged in water (Gibson, 1966). In both cases, information is inadequate because the normal constraint of the linear propagation of light is broken due to reflection or refraction. Constraints can also be artificially broken, such as when creating a "swinging room" (Lee & Aronson, 1974) in which the normally prevailing coupling between the ground and the walls is de-coupled, or an Ames' distorted room which is deliberately constructed to yield (deceptive) specificity of a normal rectangular room when viewed from one particular position (cf. Runeson, 1988). Because the perceptual system relies on the normally prevailing ecological constraints, such constructions result in anomalous and in some sense erroneous2 perception (cf. Jacobs, Runeson, & Andersson, 2001). However, errors may occur despite that both the physiological processes and available information are adequate. In some cases, observers rely on variables that do not specify a specific environmental property although a specifying variable is available (cf. Jacobs & Michaels, 2002; Withagen, 2004). Such variables may be more or less related to the property but lacks the one-to-one correspondence that the specifying variable has, that is, "a single value of the variable can go together with several states of the property" (Jacobs & Michaels, 2001, p. 564). Runeson (1989) referred to such 2. Some researchers in the Gibsonian tradition claim that "errors" and "misperceptions" do not occur if one considers evolution and the learning history of the perceiver (e.g., Michaels & Carello, 1981, cf., Cutting, 1982). 9.

(202) variables as incomplete invariants. An incomplete invariant is an informational variable that is granted by constraints that do not apply throughout the relevant environment, leading to a less than perfect correlation between the variable and a specific environmental property. For a subset of cases, or in a local region of the environment, an incomplete invariant may provide specificity however, and thus become a complete invariant (i.e., a specifying variable) in this limited region (e.g., Jacobs, Runeson, & Michaels, 2001; Runeson, 1989). Jacobs and Michaels (2001) used the term nonspecifying variables for incomplete invariants, which also is the term used in this thesis. Why would nonspecifying variables be used when they lead to less than perfect perceptual achievements or to misperceptions and errors? As pointed out by Withagen (2004), certain relevant environmental properties might not be specified by informational variables. In such cases, perceivers have no option but to pick up variables that are non-specific to the property. In cases when specifying information exists, it may nevertheless be non-exploited due to an unadapted perceptual system. Moreover, because nonspecifying variables may be quite informative, such variables may be picked up because they lead to “fairly accurate” perception (Withagen, 2004, p. 242). Gibson (1966) pointed at the economy in perception, "the information registered about objects and events becomes only what is needed, not all that could be obtained" (p. 286). However, with extended exposure or practice, perceivers may learn to pick up variables that are more informative, and thus improve in their perceptions.. Perceptual learning In the direct-perception view, perceptual learning is about differentiation of previously undifferentiated stimuli, or about changes in what variable in stimulation that is exploited.3 An example of the former is a wine connoisseur that has learned to differentiate variables of chemical stimulation, which for the novice are undifferentiated (Gibson & Gibson, 1955). Examples of the latter are found in learning studies concerning visual perception of the pulling force of a stick-figure (e.g., Michaels & de Vries, 1998), length of wielded non-seen objects (e.g., Michaels, Arzamarski, Isenhower, & Jacobs, 2008; Wagman, Shockley, Riley, & Turvey, 2001), visually guided braking (Fajen & Devaney, 2006), and visual discrimination of relative mass of colliding objects (e.g., Jacobs et al., 2000; Jacobs, Runeson, & Michaels, 2001; Runeson et al., 2000). In these studies, observers changed in what variable they used for their judgments. As novices, they generally relied on nonspecifying variables, however not always the same for all individuals (Jacobs & 3 Yet another type of learning is calibration, which concerns scaling of perception or action relative the exploited informational variable (e.g., Bingham & Pagano, 1998; Withagen & Michaels, 2004).. 10.

(203) Michaels, 2001; Jacobs et al., 2000; Jacobs, Runeson, & Michaels, 2001; Michaels & de Vries, 1998; Withagen & Michaels, 2005). During practice, observers learned to detect the specifying variable, or a nonspecifying variable that correlated higher with the property that they were to judge than the initial variable did. That is, with practice, generally observers converge on relying on variables that are more informative. Perceptual learning due to convergence on more informative variables is referred to as the education of attention. Attention is what controls detection of information and enables noticing of critical differences and less noticing of irrelevancies (Michaels & Carello, 1981; Gibson, 1966, p. 52). The environment influences attention, "interesting bits of structure, particularly motions, draw the foveas toward them" (Gibson, 1966, p. 260). However, attention is also influenced by observers' intentions (Shaw & Kinsella-Shaw, 1988). Perceptual learning will then also involve the education of intention (Jacobs & Michaels, 2002). With practice, perceivers will learn and improve in which perception they will intend to actualize. This will partially determine where attention will be directed and thus what variables will be exploited. As pointed out by Jacobs and Michaels (2002), intentions constrain task situations and, indirectly, intentions determine which is the specifying variable in a particular situation. "Given intentions, some variables can be said to be information in a specificational sense and others cannot." (p. 134) In perception experiments, it is generally assumed that participants intend to perceive the property they are instructed to judge. Above, perceptual learning and the education of attention were said to be influenced both by the environment and by the perceiver’s intentions. However, the earlier ecological theories were not explicit about how learning would come about (cf. Michaels & Beek, 1995). The recent direct learning model (Jacobs & Michaels, 2007; Michaels et al., 2008) suggests that learning is guided by convergence information. Convergence information consists of a “higher-order relation of detectable quantities”, defined over a time interval or over a number of trials. The higher-order relation would involve certain properties of the environment, perceptual judgments, and feedback. Over time, the perceiver would detect convergence information which would “push” him or her to attend to more useful informational variables, ultimately the specifying variable. Notably, no mediating "cognitive" processes are needed for learning to occur.. 11.

(204) Indirect perception The indirect4 approach to perception belongs to a tradition older than the direct-perception approach. The major differences between the approaches concern how the stimulation is portrayed, and the involvement of mediating processes between stimulation and percepts. Moreover, whereas in the direct view perception is about "keeping-in-touch with the world" (Gibson, 1979, p. 239), perception in the indirect views is about creating a mental representation of the environment (Epstein, 1995). In the indirect views, stimulation is seen as ambiguous with respect to the environment. A given event or property in the environment would give rise to a specific retinal pattern, but a given retinal pattern could be caused by numerous distal events or properties; the inverse projection problem (see Epstein, 1995). To solve the inverse projection problem, and thus to become informed about the environment, the perceiver must somehow disambiguate the sensations. Often, the problem becomes how to construct a threedimensional representation from the two-dimensional retinal image. Indirect theories mainly focus on how this disambiguation and construction would occur. Helmholtz (1867/1925) suggested that the perceptual system relies on unconscious inferences to solve the inverse projection problem (see Allik & Konstabel, 2005, for a discussion of the origins of this term). The inferences are built on premises such as "under normal conditions, only 3-D structures generate patterns of binocular disparity". If retinal stimulation indeed forms a pattern of binocular disparity, the conclusion is drawn that "the object in my visual field is very likely a 3-D structure" (from Epstein, 1995). Helmholtz' idea of unconscious inferences in perception is the core in cognitive constructivist notions of perception (e.g., Bruner, 1957; Brunswik, 1956; Knill, Kersten, & Yuille, 1996; Rock, 1983). The inferences resemble the cognitive operations used for conscious problem solving (see Kubovy, Epstein, & Gepshtein, 2003). For example, in Rock's (1983) view "perception is intelligent in that it is based on operations similar to those that characterize thought" (p. 1). Generally, the inferences, and thus the perceptual outcomes, rely on internal premises, stored knowledge, or assumptions about the environment and about how stimulation is structured vis-à-vis the environment. For example, in discussing anomalous perception Rock (1983) wrote "the perceptual system 'knows' certain laws of optics that normally obtain and then 'interprets' seeming departures from these laws in such a way as to be compatible with them" (p. 10). In more recent indirect views of perception, as in the Bayesian approach (Knill et al., 1996; Kersten, Mamassian, Yuille, 2004), internal 4. Gibson (1979), however, used the term "indirect perception" for perception through pictures, x-ray images, microscopes and telescopes etc.. 12.

(205) assumptions also about the structure of the environment are central (cf. Kubovy et al., 2003). "Explicit models of world structure (i.e., regularities in properties of the world) are needed to completely characterize both the information provided in images for perception and the actual inferences made by the visual system in the course of perception" (Knill et al., 1996, p. 2). According to the Bayesian model of perception, the perceptual outcome is in the form of a probability distribution, which is determined in part by a coding scheme based on the physics of light reflection, refraction etc. and signal corruption, and in part by the statistical structure of the world (the "prior probability distribution"). One example of such a "prior" is the assumption that light is coming from above (Kersten et al., 2004, p. 283). According to the computational constructivist approach (e.g., Marr, 1982), perception does rely neither on stored knowledge and assumptions, nor on cognitive operations. Instead, computations in the visual system embody certain "natural constraints" in the interpretation of the retinal image. Examples of such natural constraints are that matter is predominately coherent, that surfaces of objects most often are rigid, and that substances tend to be opaque (Pylyshyn, 1999, p. 355). These environmental regularities have acted to shape computational modules in the visual system, "the visual system does not need to access an explicit encoding of the constraint: it simply does what it is wired to do" (Pylyshyn, 1999, p. 354). The premises, assumptions, and "natural constraints" in the indirect theories resemble, and seem to have the same function as, the global and ecological constraints in the direct-perception model. A major difference is, however, that in the indirect views those elements are said to reside within the organism, whereas in the direct view the constraints reside outside the organism, and the consequences of which are exploited by the perceptual system (cf. Jacobs, Runeson, & Andersson, 2001; Kubovy & Epstein, 2001; in response to Shepard, 2001). There have been attempts to reconcile the indirect and direct views. For instance, Braunstein (2002) pointed out that the term "inference" in the indirect theories should not necessarily be taken to imply that perception is intelligent and thought-like. Instead, inferential mechanisms should be seen as instantiating rules (which is explicit in the computational constructivist models); thereby they will constitute smart mechanisms (Runeson, 1977). "If the concept of inference is completely separated from intelligence, thought, and active use of knowledge, and is allowed to encompass smart mechanisms and resonance, there is no need for direct perception theorists to object to inference." (Braunstein, 2002, p. 99) (See also Hatfield, 1988, 1990; Norman, 2002; Wagemans, 1990.) It can also be noted that proponents of indirect perception usually hold that something is "directly perceived". In an article that criticized the direct-perception approach, Fodor and Pylyshyn (1981) noted that “even theories that hold that perception of many properties is inferentially mediated must assume that the detection of some properties is 13.

(206) direct (in the sense of not inferentially mediated). Fundamentally, this is because inferences are processes in which one belief causes another. Unless some beliefs are fixed in some way other than by inference, it is hard to see how the inferential processes could get started.“ (p. 155) (see also Michaels & Carello, 1981, p. 183).. Modes of apprehension and the mode-transition model Runeson et al. (2000) proposed that humans function in either of two qualitatively different modes when apprehending the environment. One is the perceptual mode, which entails pick up of information that directly leads to judgments or actions. The other is the inferential mode, which entails pick up of cues or clues that via rules or inferences lead to judgments or actions. The mode-transition model of perceptual skill acquisition is a complement to direct perception notions of perceptual learning, as discussed above. Perceptual skill acquisition would proceed as a transition from the inferential to the perceptual mode, together with a shift from low- to high-quality information usage. "True beginners start out in an inferential-reasoning mode, in which they are groping for an adequate way to master the task. Sooner or later they discover a direct 'intuitive' mode of attention" (Runeson et al., 2000, p. 550). For novices, the inferential mode would provide "a pedagogical vehicle with which observers explore the task situation and precipitate feedback, allowing their perceptual system to discover new possibilities" (p. 550). The modetransition model thus differs from the direct learning model in that inferences play a role in the learning process, and thus that learning is not solely guided by information. However, as indicated by the quote above, not only inferences would influence the skill acquisition process. One essential part of perceptual skill acquisition would be perceptual “discovery”, that is, detection of informational variables. The mode distinction sprang out of a debate between proponents of the direct-perception approach (Runeson, 1995; Runeson & Vedeler, 1993) and proponents of a cue-heuristic approach (Gilden, 1991; Gilden & Proffitt, 1989, 1994; Proffitt & Gilden, 1989) regarding the informational basis in dynamic-event perception. Gilden and Proffitt argued that only simple properties of events, such as directions and speeds, are available for perception, whereas “multidimensional” properties, such as velocity-vector components in collisions, are not. Thus, when asked to judge the relative mass of objects involved in collisions, observers are left with impressions of speed and directions of the objects' trajectories. Observers then have to resort to heuristics, one for each category of information, to make their judgments. Runeson et al. (2000; Runeson & Vedeler, 1993) showed that Gilden and Proffits' cue-heuristic explanation applied only to novices. After practice, observers. 14.

(207) performed better than expected from Gilden and Proffitt's model, and their performance fitted better to the mass ratio invariant model.. What are the inferences in the inferential mode of apprehension? What would characterize the inferential mode? Runeson et al.'s (2000) article was directed mainly to a cue-heuristically inclined audience, and aimed to show that visual pick up of specificational information about dynamics does in fact occur for experienced perceivers (Runeson, personal communication, 2009). However, they did not go into details of what inferential-mode functioning would consist of. Instead, it seems they accepted Gilden and Proffitt’s (1989) model of dynamic event perception as being an example of novices' inferential-mode functioning, "the general character of Gilden and Proffitt's cue-heuristic model may be useful in describing the explorative endeavors of beginners on unfamiliar tasks" (Runeson et al., 2000, p. 550). According to Gilden and Proffitt’s cue model, people do have some “common-sense notions about how the world operates”, and they make dynamic judgments "on the basis of heuristics that articulate these notions" (Gilden & Proffitt, 1989, p. 374). According to Gilden (1991) "the mapping [between kinematics and dynamics] that people apparently use is based on a small set of separate ideas that they have about the way the world works." (p. 557) When relying on Gilden and Proffitt's model as describing inferential mode-functioning, it seems that the heuristics, or inferences, used in the inferential mode would be deliberately used rules. The inferences would thus not consist of the “unconscious inferences” that are the essential component of indirect notions of perception. According to Hecht (1996) however, the heuristics in Gilden and Proffitt's model would indeed be "unconscious perceptual heuristics" (p. 65). Hecht (2000) stated that "the term heuristic is suggestive of rules being consciously applied by the perceiver or by some homunculus looking at the retinal image, but this it not meant. Perceptual heuristics . . . are usually not available to introspection." (p. 4) Thus, Gilden and Proffitt's model seems to fall under the indirect notion of perception. The approach taken in this thesis is that the inferences in the inferential mode are not unconscious inferences, or "perceptual heuristics", carried out in the perceptual system. Instead, the inferences would be deliberately and consciously applied rules or heuristics.5 Inferential-mode functioning entails pick up of nonspecifying variables, which act as cues or clues, enabling inferring of unknown properties from the perceptually known. To function inferentially, observers need to have, or to form, an idea about how the non5. Hence, the introspection-based indicators of the modes that will be employed in the thesis (realism of confidence and mode reports) are not suited for testing Gilden and Proffitt's (1989) model, providing that their "perceptual heuristics" are unavailable for introspection.. 15.

(208) specifying variables are related to the sought-after property. Perceptualmode functioning, in contrast, entails pick up and usage of variables that specify, or that from the observer's point of view seem to specify, the soughtafter property. Because the information seems to specify the sought-after property, it is directly used or acted upon. Thus, in both modes, pick up of information would occur, but the difference is in what kind of information is obtained and, more critically, how observers use information; directly leading to judgments or leading to judgments via rules. Moreover, the inferential mode is not a case of perception as it is conceptualized by the indirect theories, and the mode distinction is not a distinction between direct and indirect perception.. Mode transition: a shift from "thinking" to "perceiving" As noted above, the mode-transition model of perceptual skill acquisition involves both a change in the effective variable for perception and a change in the mode of apprehension. Runeson et al. (2000) pointed to similarities between the mode-transition model and models of skill acquisition in motor tasks. In motor tasks, skill acquisition would proceed by the actor "freezing" all but a few of the bodily degrees of freedom, which allows deliberate control of the remaining parts (e.g., Bernstein, 1967; Fowler & Turvey, 1978; Vereijken, van Emmerik, Whiting, & Newell, 1992). Although the resulting movements are inefficient, those movements help reveal the dynamic characteristics of the action system, which in turn allows the actor or the motor system to discover a task-specific way of coordination. Analogous to the freezing of bodily degrees of freedom, the inferential use of cues would capture only limited features of the available information. Use of cues would lead to "a suggestive pattern of success and failure", which, in turn, would allow the perceptual system to discover "a stable, qualitatively different, and more efficient mode of cognizing" (Runeson et al., 2000, p. 550). Thus, the mode transition entails perceptual discovery of task-relevant information, after a phase of deliberate exploration of the task environment. The notion of perceptual skill acquisition as a change from consciously controlled performance to performance less directly subject to conscious control surfaces also in other views on skill acquisition. For example, in Anderson's ACT model (1982), cognitive skill acquisition proceeds as a transition from a declarative to a procedural stage, in which knowledge is gradually converted from declarative to procedural form. Declarative knowledge is a set of facts about the skill that are kept in working memory, and that often is verbally mediated. Procedural knowledge consists of procedures or skills specific to the task at hand, which directly apply the knowledge. (See also Fitts & Posner, 1967; Schneider & Shiffrin, 1977, for similar models of skill acquisition in visuo-motor tasks and in visual search, respectively.) The mode-transition model also resembles Dreyfus and Dreyfus 16.

(209) (1986) view on the development of expertise in that both involve changes in what information the performer considers; from limited features to more holistic information. Moreover, behavior changes from being deliberate and rule-based, to being "intuitive" requiring no awareness. Functioning in the perceptual mode, like the expert functioning of Dreyfus and Dreyfus, would be "without awareness of how it comes about, relying on the inherent sophistication of one's action or perception system" (Runeson et al. 2000, p. 551). Runeson et al.'s (2000) distinction between two modes of apprehension also resembles Heft's (1993) distinction between perceptual and analytical judgments. Heft asked observers to estimate whether a seen object was within reach or not. When the estimates were made the focal task, observers overestimated their reach, whereas when the estimates were made a subsidiary part of another task, they became very accurate. Heft suggested that observers adopted an "analytical attitude" in the focal condition, which interfered with perceptual processes. He pointed out that "psychological processes can be conceptualized as modes of adaptive functioning in relation to the ongoing flow of events…Among these psychological processes is the capacity to 'step outside' of this flow" (p. 259). The analytical attitude and the ability to step out of a flow of events would also characterize the inferential mode of apprehension. Brunswik made a distinction between intuitive and analytical modes of cognition, or between "perception" and "thinking" (1955, 1956, 1966; Hammond, 1966, p. 47). In contrast to Gibson's direct-perception approach, according to Brunswik not only thinking, but also perception relies on cues and inferences (for a comparison of Gibson's and Brunswik's approaches, see Vicente, 2003). "Perception and thinking … emerge as different forms of imperfect inferences regarding the environment, subsumable to a common behavior model patterned upon reasoning" (Brunswik, 1955, p. 108). Perception would rely on integration of many cues, which would lead to a "soft and smudgy" error profile, with a relative freedom of large errors. Thinking, in contrast, would be more "single-track", and lead to a large proportion of perfect judgments, with no near-misses, but also to occasional large or even "bizarre" judgments (Brunswik, 1966, p. 489; see also Juslin & Olsson, 1997, for a similar distinction between Thurstonian and Brunswikian origins of error). Although Brunswik's view on perception differs from the directperception approach, his predictions regarding the error profile for perception and thinking might be relevant also for distinguishing between the two modes of apprehension. Before examining this possibility, and other potential indicators of the modes, the aims of the thesis and the experimental paradigm employed are presented.. 17.

(210) Aims of the thesis. The main aims were to examine the mode-transition model of perceptual skill acquisition in discrimination of relative mass of colliding objects (Runeson et al., 2000) and to further explore possible methods to empirically distinguish the modes of apprehension. Do people change from inferential to perceptual functioning during practice? How can we know whether such a change occurred? These questions were addressed in a series of four studies, each with specific goals: Study I served as a prelude, focusing on variable usage and how it is affected by feedback. During practice, do observers change in the variable relied upon when judging relative mass? Would such a change be entirely feedback dependent? Study II tested the mode-transition model and whether the modes were related to which variables observers used. During practice, do observers change from inferential to perceptual functioning? Are nonspecifying variables used in the inferential or the perceptual mode? Realism of confidence was employed as mode indicator. Study III investigated whether the modes could be distinguished on the basis of converging evidence from introspective mode reports, event-related potentials, and response times. Are those measures congruent? Do they indicate a transition in mode? Study IV tested whether introspective mode reports and realism of confidence are congruent mode indicators.. 18.

(211) Method. Experimental paradigm The experimental paradigm involved simulated two-dimensional collisions between two circular objects. In most experiments, participants’ task was to indicate which of the objects was the heaviest in each trial. This paradigm is useful for studying the modes of apprehension for several reasons: First, the collisions resemble naturally occurring events because the movement patterns (kinematics) were generated to specify distal sources (dynamics), as granted by the laws of physics (cf. Stoffregen, 1993, on "natural" displays). This is different from paradigms in which such constraints are broken (e.g., Twardy & Bingham, 1999), and tasks with more artificial or unrepresentative stimulus materials. Thus, with the collision displays, observers would presumably not experience perceptual anomalies, and there is potential for KSD-based perception of relative mass. Second, the kinematic properties of the collisions are possible to exactly quantify, which may be harder to do in other "natural" paradigms as for instance in tasks using human point-light displays (e.g., Runeson & Frykholm, 1983). Third, the collisions contain several candidate kinematic variables that observes may use for their judgments. Fourth, presumably the task is new to the observers. Thus, in the course of practice, skill acquisition and possible accompanying mode transitions could be examined.. Candidate informational variables Figure 1 shows a graphic velocity-vector description of a collision event. During the whole event, both objects project constant, but different, speeds (SA and SB) along the sweep axis. At the moment of collision, the objects' motions change along the collision axis (wA and wB). The ratio | wA | / | wB | is proportional to the ratio of the objects’ masses, mB / mA. This mass ratio invariant thus specifies the relative mass, and is an example of KSD-based information (Runeson, 1977/1983; Runeson & Frykholm, 1983).. 19.

(212) Figure 1. Velocity-vector description of a collision. From Runeson, S., & Andersson, I. E. K. (2007). Achievement of specificational information usage with true and false feedback in learning a visual relative-mass discrimination task. Journal of Experimental Psychology: Human Perception and Performance, 33, 163–182. Reprinted by permission of the publisher. No further reproduction or distribution is permitted without written permission from the American Psychological Association.. The collisions also contain kinematic variables that do not specify the relative mass. The nonspecifying variables that were considered in the studies were the Exit-Speed variable (the ratio of the exit speeds, | vA | / | vB |), the Sweep-Speed variable (the ratio of the sweep speeds, | SA | / | SB |), and the Scatter-Angle variable (the difference in scatter angles, A - B). According to Gilden and Proffitt's (1989, 1994) cue-heuristic model, the mass ratio invariant is not available for perceptual pick up. Instead, observers would only pick up relative exit speeds or differences in scatter angles in order to judge the relative mass. The relative exit speed would be used according to the rule that the object that moves the faster after impact is the lighter. The difference in scatter angles would be used according to the rule that the object that changes direction more is the lighter. In the studies in this thesis, the Exit-Speed and Scatter-Angle variables are taken to be used according to those rules. The Sweep-Speed variable is taken to be used as if the overall faster object is the lighter. Consequently, for assessing variable usage, and 20.

(213) when scoring performance against the nonspecifying variables (Study II), responses were coded as correct when the respective rules were followed. Depending on the specific set of collisions, the Exit-Speed and ScatterAngle variables were more or less correlated with mass ratio, but always lower for Exit-Speed. For the sets of collisions in Study I, II, IV:1, and 2, Pearson product-moment correlation was .15 for Exit-Speed, and .54 for Scatter-Angle.6 The Sweep-Speed variable had zero correlation with mass ratio; it was included in Study I as a “nonsense” variable. In Study III, Scatter-Angle was nullified in test blocks (the objects changed direction to the same extent), but it was available in practice blocks, with .71 correlation with mass ratio. Exit-Speed had .48 correlation with mass ratio in test blocks, and .32 in practice blocks. As an example of a set of collisions, Figure 2 shows a graphic vector description of the test block collisions in Study III.. Figure 2. Test-block collisions in Study III. Each column contains collisions with the same mass ratio, with mass ratios given below. ExL indicate large exit speed ratios, ExM medium, and ExS small exit speed ratios.. 6. Correlations presented in Study I (p. 167) were obtained through regression analysis with a no-constant model. The Exit-Speed correlation value was lower in that calculation as compared to the correlation values presented here, and in Study II and IV. However, the sets of collisions were identical in the studies, and the respective correlations are merely alternative ways of describing the stimulus material. The ways of calculating the correlations do not affect the results and conclusions.. 21.

(214) Equipment and simulations Two-object collisions were simulated and displayed by means of an inhouse analog computer system. Analog computation represents variables by continuous voltages instead of numbers, thus providing smooth movements. The collisions were rendered on a high-intensity CRT monitor oscilloscope screen and back-projected onto a circular ground-glass screen, which in turn was viewed by the participants through a collimator lens system (for details about the equipment see Runeson et al., 2000; Runeson & Vedeler, 1993). The two objects simultaneously started to move towards each other, from various off-center positions. The motions started with smooth accelerations (natural starts, Runeson, 1974), thus there was no artifactual suddenness of the starting motions. The objects had different driving forces, yielding different incoming speeds. After 1.5 seconds they collided near the center of the screen. The impact phase, which lasted about 100 ms, was simulated with internal dynamics, leading to a brief compression of the objects in the collision dimension. After the impact phase, the objects moved along their new trajectories for 1- 3 seconds before disappearing beyond the screen or until shut off. The objects were recognized by a continuous and a broken outline, respectively, which were randomly assigned in each trial. The collisions were presented in one of 24 orientations, randomly assigned in each trial. A digital computer controlled the experimental variables, feedback, and response recording. In Study III, an additional system was used for EEG recordings. For indicating which object was the heaviest, there were two buttons beneath the screen. One button had a serrated raised edge, the other a smooth raised edge, corresponding to the outlines of the objects. For confidence ratings (Study II and IV), there was a display beneath the screen showing a 40-steps confidence scale, ranging from 50 % (meaning "I am purely guessing") to 100 % ("I am absolutely sure"). Participants turned a knob to adjust a marker on the scale to indicate their confidence. At the beginning of each trial, the marker was placed at random positions on the scale. For mode reports (Study III and IV), the endpoints of the scale represented SAW and INFERRED, respectively, and the knob was turned until either SAW or INFERRED was read beneath the scale.. Tentative indicators of mode of apprehension One aim of the thesis was to explore possible methods for empirical distinction of the modes of apprehension. Several tentative mode indicators were used: variable usage, realism of confidence (cf. Runeson et al., 2000), phenomenological mode reports (cf. Kreegipuu & Runeson, 1999, 2000), event-related potentials, and response times. 22.

(215) Variable usage Participants' variable usage was assessed over blocks of trials. For each participant and block, performance was fitted to each of the candidate variables (Mass-Ratio, Exit-Speed, Scatter-Angle, and in Study I also SweepSpeed) by means of regression analyses with a no-constant PROBIT model. As goodness-of-fit measure, McFadden’s R2 (Amemiya, 1981; referred to below as R2MF) was used. Unlike ordinary R2 measures, the R2MF indicates the fit of frequency data to a step function. With a no-constant model, the R2MF values capture the relative absence of both unsystematic (noise) and systematic (biases) errors (for details and discussion of the method, see Runeson et al., 2000). Runeson et al. (2000) did, in line with Gilden and Proffitt (1989, 1994), regard the nonspecifying variables as cues that are used inferentially, and not as variables that are used directly in a perceptual mode. Thus, according to Runeson et al., usage of nonspecifying variables would indicate inferential functioning. This has been questioned for instance by Withagen (2004; see also Hajnal, Grocki, Jacobs, Zaal, and Michaels, 2006; Jacobs, Michaels, Zaal, & Runeson, 2001; Michaels & de Vries, 1998), who argues that nonspecifying variables are used directly, and not for inferences. 7 The approach taken in this thesis is the same as Withagen's. It is acknowledged that nonspecifying variables might be picked up without being accompanied by supplementary inferential rules. This could occur for instance in the case of an unadapted perceptual system (see section above: Perceptual errors and nonspecifying variables). Hence, usage of nonspecifying variables is not taken to indicate inferential functioning. On the other hand, tentatively, usage of the specifying variable is taken to indicate perceptual functioning. As shown in Figure 1, the ratio of two velocity-change vectors specifies relative mass, and this ratio is presumably available for perceptual pick up. However, the relative mass can also be derived from an assemblage of incoming and outgoing speeds, in combination with direction changes. Hence, in principle observers could pick up those cues, combine them inferentially, and performance would still fit an invariant model. According to Runeson et al. (2000), such an explanation would be unparsimonious as compared to the single-step pick up of the invariant. Moreover, analytically complex variables, such as the mass ratio invariant, are often more useful for the perceptual system than analytically simple variables, and are likely to be picked up when available (Runeson, 1977, 1994). 7. However, the inferences that Withagen (2004) refers to are unconscious inferences that result in a representation, as in the indirect notions of perception (see above). As defined in this thesis, the inferences in the inferential mode would not be such unconscious inferences, but instead deliberately, or consciously, used rules. In this perspective, the criticism from Withagen (that usage of nonspecifying variables is not related to inferential processes) is misdirected.. 23.

(216) Following these arguments, performance characterized by usage of the mass ratio invariant indicates perceptual functioning (this is further discussed in Discussion: Realism of confidence). In sum, performance characteristics may reveal whether observers use specifying or nonspecifying variables. But variable usage, in turn, only provides at most a one-way differentiation of the modes. Nonspecifying variables might be used either in the inferential mode, as was assumed by Runeson et al. (2000), or in the perceptual mode, as was argued above, thus usage of nonspecifying variables does not indicate mode.8 However, as argued by Runeson et al., usage of the mass ratio invariant indicates perceptual-mode functioning.. Realism of confidence Realism of confidence concerns people's trust in their judgments. It is measured as the difference between the mean of participants’ confidence in their judgments (typically given as a percentage) and their actual percentage correct judgments. If the difference is positive, participants are overconfident, if it is negative, they are underconfident. In research on uncertainty in judgments, some have found that in cognitive tasks, such as general knowledge tasks or problem-solving tasks, people are overconfident (e.g., McClelland & Bolger, 1994; Yates, 1990). Others have found that people are well calibrated in such tasks, providing that test items are representative of the natural environment (e.g., Gigerenzer, Hoffrage, & Kleinbölting, 1991; Juslin, 1994). For sensory discrimination tasks, such as in discrimination of line lengths, there are findings of underconfidence (Björkman, Juslin & Winman, 1993; Juslin & Olsson, 1997; Juslin, Olsson & Winman, 1998; Olsson & Winman, 1996; Stankov, 1998), good calibration (e.g., Baranski & Petrusic, 1994, 1999), and even overconfidence (Kvidera & Koutstaal, 2008). Some researchers argue that participants' judgment confidence in cognitive and sensory tasks originates in the same type of process, and that there is no systematic difference in over/underconfidence between those tasks (Baranski & Petrusic, 1994, 1999; Ferrell, 1995; Suantak, Bolger, & Ferrell, 1996). Others, however, argue that confidence in cognitive and sensory tasks originates in different sources of uncertainty (Björkman et al., 1993; Juslin, Olsson, & Winman, 1998; Olsson & Winman, 1996). According to Juslin and Olsson’s (1997) sensory sampling model, in binary sensory tasks, confidence ratings are based on the experienced variability in a sample of impressions. The judgments, however, are based on a statistical aggregate (the central tendency in the sample). Judgments would then on average be more ac8. Hence, the error profile (Brunswik, 1956, 1966; see Introduction: Mode transition: a shift from "thinking to "perceiving") would not indicate mode.. 24.

(217) curate than suggested by the confidence ratings, which would result in underconfidence. In cognitive tasks, in contrast, confidence ratings would reflect the apparent validity of used cues: When people have appropriate knowledge of the cue validities, and with representative samples of test items, good calibration is predicted (Gigerenzer et al., 1991; Juslin, 1994). Based on Juslin and Olsson’s (1997; Juslin, Olsson, & Winman, 1998) idea that there are different sources of uncertainty in the two types of tasks, and that over/underconfidence levels thus would differ between the tasks, Runeson et al. (2000) reasoned that realism of confidence could be used as an indicator of participants’ mode of apprehension. That is, over/underconfidence levels would indicate whether participants perform a task as they would perform a cognitive or a sensory task, respectively. Underconfidence would indicate perceptual functioning, whereas overconfidence or good calibration would indicate inferential functioning. In this thesis, this idea is referred to as the mode-confidence hypothesis. Also mode transitions would be reflected in over/underconfidence levels. If observers change from relying on cues and inferences, to relying on a perceptual impression of the sought-after-property, then they would change from being overconfident or well calibrated, to being underconfident. Runeson et al. (2000) found that novice observers were well calibrated (0 % over/underconfidence with 80 % correct judgments), whereas after practice with feedback they were underconfident (-11 % underconfidence with 88 % correct). This result, together with the finding of a shift in variable usage from the Exit-Speed variable to the mass ratio invariant, suggested that a mode shift had occurred during practice. What complicates such an interpretation of over/underconfidence levels is that easy tasks (tasks that yield a large proportion correct) tend to generate underconfidence, whereas difficult tasks (tasks that yield a small proportion correct) tend to generate overconfidence, at least in cognitive tasks: the hard-easy effect (Lichtenstein & Fischhoff, 1977). According to Juslin et al. (1998), in sensory tasks there is a disposition towards underconfidence at most difficulty levels (but for a contrary view see for example Baranski & Petrusic, 1999), but the hard-easy effect will still influence the over/underconfidence scores. The hard-easy effect seems to stem from methodological artifacts such as confidence-scale end effects and regression effects, and, at least in cognitive tasks, it is considered neutral at 75 % correct (Erev, Wallsten, Budescu, 1994; Juslin, Winman, & Olsson, 2000; Suantak, Bolger, & Ferrell, 1996). In Runeson et al. (2000), proportion correct increased from pre- to posttest, which led to reduced over/underconfidence scores due to the hard-easy effect. However, when comparing over/underconfidence scores for subsets of collisions with similar proportions correct, over/underconfidence scores were still lower in the posttest. This showed that the overall lower over/underconfidence score in the posttest was not due only to a hard-easy effect, and that there was a true differ25.

(218) ence in over/underconfidence level, which strengthens the conclusion that a mode transition had occurred. Hajnal et al. (2006) criticized Runeson et al's (2000) conclusion, claiming that the difference in over/underconfidence scores emerged because participants used different variables before and after practice. Before practice, when the participants were using a nonspecifying variable, proportion correct was smaller than after practice when they had changed to using the mass ratio invariant. Since over/underconfidence is calculated as the difference between the mean of the confidence ratings and the proportion of correct judgments, the smaller proportion correct before practice would lead to an elevated over/underconfidence score. Hence, if participants were using nonspecifying variables in the perceptual mode, good calibration could still appear due to a mismatch between the experimenters' criterion of performance, and the variable actually used. In Study II, individuals’ variable usage was assessed block-wise, by their 2 R MF values for each candidate variable. Both performance and over/underconfidence were individually scored against the most used variable in each block, a procedure suggested by Hajnal et al. (2006). As was argued in Study II, this procedure would disclose possible underconfidence for usage of nonspecifying variables, which in turn would reveal whether nonspecifying variables were used in the inferential or the perceptual mode. To counteract the hard-easy effect, main comparisons in Study II, and also in Study IV, were made on subsets of collisions that yielded around 75 % correct, or between subsets of collisions with similar proportion correct.. Phenomenological mode reports In Study III, IV:1 and 2, phenomenological mode reports were collected. After each mass ratio judgment, participants indicated what mode they were functioning in; whether they "SAW" or "INFERRED" the relative mass. Unlike over/underconfidence scores, which are based on aggregated data, this method would, if successful, capture trial-by-trial mode shifts. Kreegipuu and Runeson (2000; see also Kreegipuu & Runeson, 1999) collected similar mode reports. Performance in the relative mass discrimination task improved after practice, and the proportion of SAW trials increased from 53 % to 65 %. It was concluded that it was "higher incidence" of perceptual-mode functioning after practice. Moreover, most obvious in the posttest, the proportion of SAW trials was largest for the largest mass ratios, and smallest for mass ratios near unity. Such distribution of SAW responses was expected because large mass ratios are presumably easier to discriminate than small ratios. If small mass ratios are harder to discriminate, possibly such cases would evoke cue-based judgments and thus being performed in the inferential mode (Runeson et al., 2000).. 26.

(219) Interpretation of introspective data is not unproblematic however. A fundamental question is whether the phenomena asked about are introspectively available. Nisbett and Wilson (1977) claimed that reports about higher cognitive processes are largely invalid because people do not have introspective access to such processes. Instead, introspective reports would be based on participants' "a priori theories about the causal connection between stimuli and response" (p. 233). However, unlike the cognitive processes per se, the products of those processes would be available to introspection. Nisbett and Wilson's proposal has been criticized on several grounds, as for instance how to distinguish process from product (e.g., Ericsson & Simon, 1980; White, 1988). Moreover, as demonstrated by Kellogg (1982), if cognitive processes are explicit, or consciously controlled, they may be available to introspection, in contrast to unconscious automatic processes (cf. Ericsson & Simon, 1980, p. 225). Wilson (2002) concluded that when responses are caused by the "conscious self", instead of the "adaptive unconscious", people do have at least some introspective access to the processes leading to the responses. Would it be possible to report on the modes of apprehension in the relative mass discrimination task? Perceptual-mode functioning is assumed to be phenomenologically characterized by a perceptual impression of the relative mass. Inferential mode functioning, in contrast, is assumed to be characterized by a search for, or identification of, cue variables, as for instance the objects' relative speed, and by the consciously controlled use of heuristic rules. Thus, participants were asked to introspect and report on the occurrence or nonoccurrence of a specific perceptual impression, together with the occurrence or nonoccurrence of explicit application of rules. If these characterizations of mode functioning hold, reasonably valid mode reports may be achievable. When participants are offered fixed alternatives, such as SAW and INFERRED, for reporting on their mental processes or experiences, it is essential that those alternatives fit to the actual processes and experiential phenomena that occur (cf. Ericsson & Simon, 1980; Jack & Shallice, 2001; White, 1980). In Study III, participants were instructed to indicate SAW if they "directly saw" or "felt" which object was the heavier, if they used "a feeling" or "intuition", if they "did not know why they had an impression that the object was heavier", and if they "after perceiving the relative mass was pondering over why, or tried to rationally explain why". They were instructed to indicate INFERRED if they were "figuring out" which object was the heavier, or if they "attended to a part of the movement pattern and then used some rule for judging the relative mass". In Study IV, the description of the perceptual mode was changed to the phrase "If you had an impression, directly saw, or felt that one of the objects was heavier". The phrases "a feeling" and "intuition" were omitted. As argued in Study IV, such phrases might be ambiguous and unclear. However, it is important to note that there were no independent means to verify that the descriptions of the modes in 27.

No results found